SparkSQL在日常的数据开发过程中占据着重要的地位，面对日益复杂的需求，需要建立复杂的数据结构，在将嵌套型JavaBean和复杂数据结构如Map等注册为table，以支持化腐朽为神奇，将复杂的数据结构转化为天下大同的sql语句，使得spark更加亲民。废话不多说，还是以代码实测敬上。
前戏走起，首先创建一个简单的JavaBean
public static class Point implements Serializable {
private double x;
private double y;
public Point(double x, double y) {
this.x = x;
this.y = y;
}
public Point() {
}
public double getX() {
return x;
}
public void setX(double x) {
this.x = x;
}
public double getY() {
return y;
}
public void setY(double y) {
this.y = y;
}
}
创建嵌套型的JavaBean
public static class Segment implements Serializable {
private Point from;
private Point to;
public Segment(Point from, Point to) {
this.from = from;
this.to = to;
}
public Segment() {
}
public Point getFrom() {
return from;
}
public void setFrom(Point from) {
this.from = from;
}
public Point getTo() {
return to;
}
public void setTo(Point to) {
this.to = to;
}
}
创建数组类型的JavaBean
public static class Line implements Serializable {
private String name;
private Point[] points;
public Line(String name, Point[] points) {
this.name = name;
this.points = points;
}
public Line() {
}
public String getName() {
return name;
}
public void setName(String name) {
this.name = name;
}
public Point[] getPoints() {
return points;
}
public void setPoints(Point[] points) {
this.points = points;
}
}
创建复杂数据类型的JavaBean
public static class NamedPoints implements Serializable {
private String name;
private Map<String, Point> pointMap;
public NamedPoints() {
}
public NamedPoints(String name, Map<String, Point> pointMap) {
this.name = name;
this.pointMap = pointMap;
}
public String getName() {
return name;
}
public void setName(String name) {
this.name = name;
}
public Map<String, Point> getPointMap() {
return pointMap;
}
public void setPointMap(Map<String, Point> pointMap) {
this.pointMap = pointMap;
}
}
前戏足矣，开始水到渠成。
public static void main(String[] args) {
SparkSession session = SparkSession.builder().appName("complex javabean to table")
.getOrCreate();
List<Segment> segments = Arrays.asList(new Segment(new Point(1.0, 2.0), new Point(3.0, 4.0)),
new Segment(new Point(5.0, 6.0), new Point(7.0, 8.0)),
new Segment(new Point(9.0, 10.0), new Point(11.0, 12.0)));
Encoder<Segment> encoder = Encoders.bean(Segment.class);
Dataset<Segment> dataset = session.createDataset(segments, encoder);
dataset.registerTempTable("segment_table");
Dataset<Row> sql = session.sql("select t.from.x from segment_table t");
sql.printSchema();
sql.show();
System.out.println("测试map");
Encoder<NamedPoints> namedPointsEncoder = Encoders.bean(NamedPoints.class);
Map<String,Point> map = new HashMap<>();
map.put("p1",new Point(0.0,1.1));
Map<String,Point> pointMap = new HashMap<>();
pointMap.put("p2",new Point(2.2,2.3));
pointMap.put("p3",new Point(2.3,2.5));
pointMap.put("p4",new Point(2.4,2.6));
List<NamedPoints> namedPoints = Arrays.asList(new NamedPoints("name_1", map), new NamedPoints("name_2", pointMap));
Dataset<NamedPoints> namedPointsDataset = session.createDataset(namedPoints, namedPointsEncoder);
System.out.println("===============================");
namedPointsDataset.show();
System.out.println("===============================");
try {
namedPointsDataset.createTempView("name_point");
Dataset<Row> rowDataset = session.sql("select * from name_point");
rowDataset.printSchema();
rowDataset.show();
System.out.println("+++++++++++++++++++++++++++++++++++++++++++++++");
Dataset<Row> rowDatasetDetail = session.sql("select t.name,t.pointMap,t.pointMap.key,t.pointMap.value.x from name_point t");
rowDatasetDetail.printSchema();
rowDatasetDetail.show();
System.out.println("++++++++++++++++++++++OVER+++++++++++++++++++++++++");
} catch (AnalysisException e) {
e.printStackTrace();
}
session.stop();
}
直接看结果吧
18/11/16 10:42:43 INFO CodeGenerator: Code generated in 220.044447 ms
root
|-- x: double (nullable = true)
18/11/16 10:42:44 INFO CodeGenerator: Code generated in 8.54563 ms
18/11/16 10:42:44 INFO CodeGenerator: Code generated in 6.818043 ms
+---+
| x|
+---+
|1.0|
|5.0|
|9.0|
+---+
测试map
18/11/16 10:42:44 INFO ContextCleaner: Cleaned accumulator 0
18/11/16 10:42:44 INFO CodeGenerator: Code generated in 56.920135 ms
===============================
18/11/16 10:42:44 INFO CodeGenerator: Code generated in 7.050247 ms
18/11/16 10:42:44 INFO CodeGenerator: Code generated in 6.498453 ms
+------+--------------------+
| name| pointMap|
+------+--------------------+
|name_1| [p1 -> [0.0, 1.1]]|
|name_2|[p2 -> [2.2, 2.3]...|
+------+--------------------+
===============================
root
|-- name: string (nullable = true)
|-- pointMap: map (nullable = true)
| |-- key: string
| |-- value: struct (valueContainsNull = true)
| | |-- x: double (nullable = false)
| | |-- y: double (nullable = false)
+------+--------------------+
| name| pointMap|
+------+--------------------+
|name_1| [p1 -> [0.0, 1.1]]|
|name_2|[p2 -> [2.2, 2.3]...|
+------+--------------------+
+++++++++++++++++++++++++++++++++++++++++++++++
root
|-- name: string (nullable = true)
|-- pointMap: map (nullable = true)
| |-- key: string
| |-- value: struct (valueContainsNull = true)
| | |-- x: double (nullable = false)
| | |-- y: double (nullable = false)
|-- key: struct (nullable = true)
| |-- x: double (nullable = false)
| |-- y: double (nullable = false)
|-- x: double (nullable = true)
18/11/16 10:42:44 INFO CodeGenerator: Code generated in 7.324709 ms
18/11/16 10:42:45 INFO CodeGenerator: Code generated in 8.556297 ms
+------+--------------------+----+----+
| name| pointMap| key| x|
+------+--------------------+----+----+
|name_1| [p1 -> [0.0, 1.1]]|null|null|
|name_2|[p2 -> [2.2, 2.3]...|null|null|
+------+--------------------+----+----+
++++++++++++++++++++++OVER+++++++++++++++++++++++++
从运行结果来看，对嵌套型JavaBean的取值，可以直接以.的形式取值。
对map类型来说，可以取出map数据结构的整体结果，但是具体的key-value取值显示都为null
或许是我的写法有问题，或许是sparkSQL在这一块还不太完善，希望能够给与建议。
第一次写博客，轻拍。。。
